Manuel Lemos wrote:
> Hello,
>
> on 10/31/2005 02:38 PM jonbouyw said the following:
>
>>Yes I did run the conformance tests and it passes all of the tests
>>including NULLS.
>
>
> I haven't tried it yet. Does it skip any tests or performs all the 16
> tests? I assume that at least auto-increment tests are skipped.

Do you have any tests for REPLACE? It should fail those as sqlite does
not return the same affected row values as mysql ..

>>I'm not clear on how the autoincrement works within Metabase at the
>>moment, but as you know SQLite will treat a primary key field of the
>>type INTEGER as an autoincrement field, I'm investigating how to
>>incorporate in the Metabase context. I would appreciate any
>>suggestions you have in this regard.
>
>
> That is very similar to MySQL. You may want to take a look at its driver
> class.

You can specific autoincrement nonetheless and IIRC it will get stored
by sqlite, which would make reverse engineering clearer.

>>Also I'm looking into the manager class as I have some ALTER TABLE
>>functions to emulate the ALTER TABLE functions that are lacking in
>>SQLite which are implemented by recreating a new table with the
>>alterations included an then re-writing the original database. This
>>is slow but does provide full ALTER TABLE functionality. Again I'm
>>investigating how to include this within the Metabase context.
>
>
> MySQL implements table alterations that way but the necessary SQL is simple.
>
> To implement a full blown AlterTable from scratch is tricky but many
> people will appreciate your effort. Just try to implement one kind of
> table alteration at a time playing with a test schema, and let me know
> when you have questions.

Sounds interesting indeed. I will probably steal with for MDB2 :-)

>>As well I expect to be able to produce a SQLite PDO driver which
>>shouldn't be too dissimilar to the SQLite one already provided.
>
>
> PDO may be more complicated. It is better to start from a generic base
> PDO driver class, like there is for ODBC, but not for PDO right now.
> Lets deal with one driver at a time.

I agree that this approach would make the most sense.

regards,
Lukas